- Improved controls for tool-calling and structured output flows.
- Reduced latency for selected model and tool combinations.
- Apps using older completions-style flows should review migration paths.
Community-rated changelogs
- Console workflows now make prompt comparison easier for app teams.
- New guardrail hints help catch risky prompt changes before release.
- Deployment views now surface slow routes and cold-start patterns.
- Teams can compare latency before and after a production rollout.
- Auth docs now include clearer examples for user-owned rows.
- Teams using custom claims should review policy snippets before copying them.
- Workers requests can be inspected with more trace context.
- Useful for debugging edge latency without adding custom logging.
- Cache access is tightened for workflows triggered by forked pull requests.
- Security posture improves, but CI time may increase for some OSS repos.
- Multi-platform builds can reuse cache layers more predictably.
- CI pipelines with ARM and x86 targets may see shorter build times.
- Docs now explain more cases where compiler adoption should be incremental.
- Teams with custom hooks should pay attention to lint and memoization notes.
- Production build fixes target route generation and CSS ordering.
- Projects with custom bundler assumptions should validate before upgrading.
- Incremental rebuilds are faster in projects with many imported CSS files.
- Configuration edge cases around content scanning were tightened.
- The beta includes narrower inference for common conditional patterns.
- Library authors should test declarations against stricter edge cases.
- Security release updates bundled dependencies and runtime patches.
- Production services should schedule patch rollout promptly.
- Client generation handles workspace output paths more consistently.
- Monorepos with custom package boundaries should re-check generated imports.
- Patch release includes planner fixes and replication reliability updates.
- DB teams should review extension compatibility before maintenance windows.
- Updated docs call out persistence settings that can surprise small teams.
- Operational guidance is clearer for cache-only versus durable workloads.
- Checkout recovery flows expose more controls for abandoned sessions.
- Billing teams can tune recovery without rebuilding custom payment pages.
- Session token refresh timing is adjusted for better reliability.
- Apps with custom middleware should test stale-session edge cases.
- Audience imports now return clearer validation errors.
- Teams can catch malformed subscriber records before sending campaigns.
- Grouping logic changes can merge or split some existing issues.
- Teams with strict alert routing should watch notification volume after rollout.
- Flag payload previews now support richer values before rollout.
- Useful for teams shipping configuration-heavy experiments.
- Pricing docs now include clearer cached-token examples.
- Teams estimating AI margins can model repeated context more accurately.
- Project transfers now show clearer warnings for domains and environment variables.
- Good quality-of-life update for agencies and multi-team workspaces.
- Image transformations can now be cached with more explicit controls.
- Apps serving many resized assets may reduce repeated processing.
- Secret scanning recognizes more provider token patterns.
- Security teams may see new alerts from historical commits.
- Docs explain stricter remote image matching with clearer examples.
- Teams debugging production image failures get a faster checklist.
- D1 projects can test a more direct backup restore workflow.
- Still beta, so teams should validate recovery paths outside production.
- Compose watch handles more host file event edge cases.
- Local development loops should be less likely to miss rebuilds.
- Auto-import suggestions rank local project exports more consistently.
- Large monorepos may see less noisy editor completions.
- Version notes call out invoice rendering differences more clearly.
- Finance-heavy apps should verify invoice previews before upgrading API versions.
- Session Replay privacy settings include reusable selector presets.
- Teams can reduce sensitive-data risk without bespoke masking config.
- Surveys can target users based on feature flag exposure.
- Useful for collecting feedback from experiment cohorts.
- Invitation templates can include more organization-specific variables.
- B2B apps get better onboarding copy without custom email plumbing.
- Console previews now show how safety settings affect request behavior before rollout.
- Teams can compare stricter defaults without rerunning the whole app flow.
- Pricing guidance now explains where batch inference improves cost efficiency.
- AI teams estimating async workloads get clearer latency and spend expectations.
- Endpoint deployments can now add guards before a model revision becomes live.
- Useful for teams serving many tenant-facing models with staged evaluation.
- Service metrics now show short-lived memory spikes instead of only averaged usage.
- Teams debugging background workers can catch noisy deploys faster.
- Deploy previews now warn earlier when preview services rely on shared database state.
- Good safety improvement for teams testing schema-affecting changes.
- New build image defaults upgrade both Node and Bun versions for fresh projects.
- Existing sites should pin versions if release timing matters for CI stability.
- Autoscaling settings add cooldown controls to reduce noisy scale-up and scale-down loops.
- Helpful for APIs with bursty traffic and tight regional budgets.
- Workspace installs now explain more hoisting conflicts instead of failing silently.
- Monorepo teams get a faster path to resolving lockfile drift and duplicate deps.
- Project settings now show a clearer audit view for environment variable changes.
- Useful for teams managing many preview deployments and shared edge apps.
- Release notes now call out when vector search index rebuilds may affect query freshness.
- AI product teams can better schedule maintenance around retrieval-heavy workloads.
- App Hosting previews add cache invalidation controls before production rollout.
- Teams can test static asset refresh behavior without guessing CDN defaults.
- Project workflows can now route new issues with project-level triage rules.
- Product teams with shared inboxes get fewer manual cleanup steps after launch days.